SIMPLELINK-CC26X2-SDK: Connection drops when limited discoverable period ends

Oh, and for what it's worth, MAX_NUM_BLE_CONNS is set to 1, though I tried setting it to 2 just to see if letting Simple Peripheral restart advertising upon GAP_LINK_ESTABLISHED_EVENT would affect this behavior; it did not.  Of course, there is also still the call to re-enable advertising in the GAP_LINK_TERMINATED_EVENT, and yet advertising never restarted after disconnect automatically either.  I just checked setting a breakpoint in GAP_LINK_TERMINATED_EVENT, and I don't ever hit it when the connection drops right when the advertising timer would have stopped.

If I change that one line of code back to 

      GapAdv_enable(advHandleLegacy, GAP_ADV_ENABLE_OPTIONS_USE_MAX , 0);

then I can stay connected indefinitely, and when I disconnect gracefully from the central, I receive GAP_LINK_TERMINATED_EVENT and advertising begins again.  So what's going wrong with limited discoverable mode?

I updated everything from 2.10, through 2.20, to 2.30 (a very painful process by the way), and I'm still seeing the exact same issue.

I believe I've found a bug in the SDK.  From simple_peripheral.c:

static void SimplePeripheral_processGapMessage(gapEventHdr_t *pMsg)
{
  switch(pMsg->opcode)
  {
...
    case GAP_LINK_ESTABLISHED_EVENT:
    {
...
      uint8_t numActive = linkDB_NumActive();
...
      if (numActive < MAX_NUM_BLE_CONNS)
      {
        // Start advertising since there is room for more connections
        GapAdv_enable(advHandleLegacy, GAP_ADV_ENABLE_OPTIONS_USE_MAX , 0);
        GapAdv_enable(advHandleLongRange, GAP_ADV_ENABLE_OPTIONS_USE_MAX , 0);
      }
      else
      {
        // Stop advertising since there is no room for more connections
        GapAdv_disable(advHandleLongRange, GAP_ADV_ENABLE_OPTIONS_USE_MAX , 0);
        GapAdv_disable(advHandleLegacy, GAP_ADV_ENABLE_OPTIONS_USE_MAX , 0);
      }

If you've maxed out your connections, GapAdv_disable is called on the advertising sets, but there are too many parameters.  See the function prototype from gap_advertiser.h:

extern bStatus_t GapAdv_disable(uint8 handle);

That's consistent with the documentation here as well: http://dev.ti.com/tirex/content/simplelink_cc26x2_sdk_2_30_00_34/docs/ble5stack/ble_user_guide/doxygen/ble/html/group___gap_adv.html#ga4e45550a6887067a26b8cc797e3d51fe.

It would seem that the extra parameters are copy/paste errors from GapAdv_enable().  But pushing extra values on the stack that the called function isn't expecting seems really bad, and possibly the source of my problem.  So I changed this in my code to the following:

        GapAdv_disable(advHandleLegacy);

Sadly, that didn't help me any.  Probably still an issue; just not mine apparently.

I did try a fresh build of Simple Peripheral on the launchpad at both 2.10 and 2.30, and I likewise see my issue present in the former, but not in the latter.  So is it something I could have carried with me when I ported the project up to the latest SDK version?  If so, since you guys fixed it and therefore hopefully know where it was, can you at least tell me where to look?

Hi,

The usage of GapAdv_disable is indeed wrong, so I would file a bug. Thanks for identifying a bug and reporting it back to us.

However, the STACK only takes the first parameter, that means it will not affect the functionality.

The reason there is no compilation error is because of icall_ble_api.h redefined the function which disable the parameter check.

#define GapAdv_disable(...)                                                             (icall_directAPI(ICALL_SERVICE_CLASS_BLE, (uint32_t) IDX_GapAdv_disable , ##__VA_ARGS__))

There has been lots of code update in the STACK so it's very difficult to identify which part of the code in the STACK makes the difference.

The quickest way for me to help you is for you to provide code snippet and I can do the testing here.

I spent a decent amount of time over the last few days trying to narrow the issue down myself.  After I stated last week that I did not see the issue on a fresh workspace/import/build of SimplePeripheral 2.30 running on the Launchpad (it's worth noting this was the first time I had used the built-in target SoC), I replaced simple_peripheral.c, simple_peripheral.h, simple_gatt_profile.c, and simple_gatt_profile.h with the profile/service files from our project with the issue.  When I ran this on the Launchpad (also the first time I've run our code on hardware that wasn't ours), the issue was still present.  Since then, I had been slowly converting everything back to the default versions, function by function, rebuilding and testing every step of the way, assuming that at some point, the issue would go away, and I'd be able to isolate the code that's causing the issue.

That seemed like a good plan, except the issue never went away.  I've now converted the entire project back to the point where a file diff utility says all the source files are identical (with the exception of the single legacy advertising set using a limited duration), and yet the issue persists.  I then created yet another fresh workspace/import/build of SimplePeripheral 2.30, changed only the advertising, and ran it on the same Launchpad; once again, it disconnected right at the end of the advertising duration timer.  It would seem that it maybe only worked the first time because it was a fresh board that hadn't gone astray yet, and once it did, I can't ever get it to go back, even with a practically unmodified example project.

Testing this theory, I borrowed my coworker's Launchpad, and with the second clean project still open, loaded the exact same output files.  It did NOT disconnect, even though this was the exact same project that had just disconnected on my Launchpad.  Still keeping the project open, I reconnected my Launchpad again, reran the debugger, loaded the project again, and it continued to have the issue.  I then went back to my coworker's Launchpad and reloaded it again, and this time, it too disconnected.

I know that's a flurry of test setups, and it might be difficult to keep it all straight, but let me call attention to the fact that my coworker's Launchpad never once had our profile/service loaded on it; the only thing I loaded onto it was the practically-out-of-the-box build of 2.30's SimplePeripheral example with only the basic change to advertising.  And with no changes to any of the files between tests, it worked the first time without disconnecting, and in a few attempts since, it has disconnected every time.

So even though you saw it work, I did too, once, and I would ask you to test it again a few times.  The extent of the changes I made to simple_peripheral.c are entirely encapsulated in the snippet below:

        // Enable legacy advertising for set #1
//        status = GapAdv_enable(advHandleLegacy, GAP_ADV_ENABLE_OPTIONS_USE_MAX , 0);
        status = GapAdv_enable(advHandleLegacy, GAP_ADV_ENABLE_OPTIONS_USE_DURATION, 4500);
        SIMPLEPERIPHERAL_ASSERT(status == SUCCESS);
/*
        // Use long range params to create long range set #2
        GapAdv_params_t advParamLongRange = GAPADV_PARAMS_AE_LONG_RANGE_CONN;

        // Create Advertisement set #2 and assign handle
        status = GapAdv_create(&SimplePeripheral_advCallback, &advParamLongRange,
                               &advHandleLongRange);
        SIMPLEPERIPHERAL_ASSERT(status == SUCCESS);

        // Load advertising data for set #2 that is statically allocated by the app
        status = GapAdv_loadByHandle(advHandleLongRange, GAP_ADV_DATA_TYPE_ADV,
                                     sizeof(advertData), advertData);
        SIMPLEPERIPHERAL_ASSERT(status == SUCCESS);

        // Load scan response data for set #2 that is statically allocated by the app
        status = GapAdv_loadByHandle(advHandleLongRange, GAP_ADV_DATA_TYPE_SCAN_RSP,
                                     sizeof(scanRspData), scanRspData);
        SIMPLEPERIPHERAL_ASSERT(status == SUCCESS);

        // Set event mask for set #2
        status = GapAdv_setEventMask(advHandleLongRange,
                                     GAP_ADV_EVT_MASK_START_AFTER_ENABLE |
                                     GAP_ADV_EVT_MASK_END_AFTER_DISABLE |
                                     GAP_ADV_EVT_MASK_SET_TERMINATED);

        // Enable long range advertising for set #2
        status = GapAdv_enable(advHandleLongRange, GAP_ADV_ENABLE_OPTIONS_USE_MAX , 0);
        SIMPLEPERIPHERAL_ASSERT(status == SUCCESS);
*/
        // Display device address
        Display_printf(dispHandle, SP_ROW_IDA, 0, "%s Addr: %s",
                       (addrMode <= ADDRMODE_RANDOM) ? "Dev" : "ID",
                       Util_convertBdAddr2Str(pPkt->devAddr));

        if (addrMode > ADDRMODE_RANDOM)
        {
          SimplePeripheral_updateRPA();

          // Create one-shot clock for RPA check event.
          Util_constructClock(&clkRpaRead, SimplePeripheral_clockHandler,
                              SP_READ_RPA_EVT_PERIOD, 0, true,
                              (UArg) &argRpaRead);
        }
      }

      break;
    }

    case GAP_LINK_ESTABLISHED_EVENT:
    {
      gapEstLinkReqEvent_t *pPkt = (gapEstLinkReqEvent_t *)pMsg;

      // Display the amount of current connections
      uint8_t numActive = linkDB_NumActive();
      Display_printf(dispHandle, SP_ROW_STATUS_2, 0, "Num Conns: %d",
                     (uint16_t)numActive);

      if (pPkt->hdr.status == SUCCESS)
      {
        // Add connection to list and start RSSI
        SimplePeripheral_addConn(pPkt->connectionHandle);

        // Display the address of this connection
        Display_printf(dispHandle, SP_ROW_STATUS_1, 0, "Connected to %s",
                       Util_convertBdAddr2Str(pPkt->devAddr));

        // Enable connection selection option
        tbm_setItemStatus(&spMenuMain, SP_ITEM_SELECT_CONN, TBM_ITEM_NONE);

        // Start Periodic Clock.
        Util_startClock(&clkPeriodic);
      }

      if (numActive < MAX_NUM_BLE_CONNS)
      {
        // Start advertising since there is room for more connections
//        GapAdv_enable(advHandleLegacy, GAP_ADV_ENABLE_OPTIONS_USE_MAX , 0);
        GapAdv_enable(advHandleLegacy, GAP_ADV_ENABLE_OPTIONS_USE_DURATION, 4500);
//        GapAdv_enable(advHandleLongRange, GAP_ADV_ENABLE_OPTIONS_USE_MAX , 0);
      }
      else
      {
        // Stop advertising since there is no room for more connections
//        GapAdv_disable(advHandleLongRange, GAP_ADV_ENABLE_OPTIONS_USE_MAX , 0);
        GapAdv_disable(advHandleLegacy, GAP_ADV_ENABLE_OPTIONS_USE_MAX , 0);
      }

      break;
    }

    case GAP_LINK_TERMINATED_EVENT:
    {
      gapTerminateLinkEvent_t *pPkt = (gapTerminateLinkEvent_t *)pMsg;

      // Display the amount of current connections
      uint8_t numActive = linkDB_NumActive();
      Display_printf(dispHandle, SP_ROW_STATUS_1, 0, "Device Disconnected!");
      Display_printf(dispHandle, SP_ROW_STATUS_2, 0, "Num Conns: %d",
                     (uint16_t)numActive);

      // Remove the connection from the list and disable RSSI if needed
      SimplePeripheral_removeConn(pPkt->connectionHandle);

      // If no active connections
      if (numActive == 0)
      {
        // Stop periodic clock
        Util_stopClock(&clkPeriodic);

        // Disable Connection Selection option
        tbm_setItemStatus(&spMenuMain, TBM_ITEM_NONE, SP_ITEM_SELECT_CONN);
      }

      // Start advertising since there is room for more connections
//      GapAdv_enable(advHandleLegacy, GAP_ADV_ENABLE_OPTIONS_USE_MAX , 0);
      GapAdv_enable(advHandleLegacy, GAP_ADV_ENABLE_OPTIONS_USE_DURATION, 4500);
//      GapAdv_enable(advHandleLongRange, GAP_ADV_ENABLE_OPTIONS_USE_MAX , 0);

Any update on this?

My colleague attempted to do this, and found an 18.9, but it wouldn't let him do 18.01.  It also resulted in a number of errors.  I will see if I can get him to share some details on this, but in the meantime, what is the best way to go about this?  I saw nothing in the documentation here that mentioned this as a requirement.  Where can we find the requirements listed, and a guide that shows the proper way to update?

The requirement is listed in release note. Release note comes with the SDK installation:

ti/simplelink_cc26x2_sdk_2_30_00_34/release_notes_simplelink_cc26x2_sdk_2_30_00_34.html

You can also use CCS 8.2 to see if it gives you less trouble.

We ran into some issues getting the latest versions of CCS and the compiler to work (https://e2e.ti.com/support/tools/ccs/f/81/t/753415, https://e2e.ti.com/support/wireless-connectivity/bluetooth/f/538/t/751100).  While Tyler was working on that, I've been focused on a different project, and haven't been able to make it back to this one yet.